Artificial Neural Network based Dual Layered Predictive Model for Rare Attack Detection
Utpal Shrivastava, Neelam Sharma · 2020 International Conference on Computational Performance Evaluation (ComPE) · 2020
Computer network is expanding day by day and number of users are increasing on the network. There are many attacks on the network with makes user on risk, leading to security of their data. Attackers are using novel method for getting information of users. Intrusion detection system (IDS) is used to detect such attack done on the network by monitoring the network traffic. In general attacks are of four types on the network in which Denial-of-Service (DoS) and Prob attacks are of majority categories (generally found or common), User-to-Root (U2R) and Remote-to-Login (R2L) are of minority categories (rarely found or rare). The rare attacks are very harmful for a host. The present day systems are not able to effectively detect such attacks. In this approach, a model is proposed of two layers to improve the detection rate of minority attacks. Feature selection for minor attacks is done using statistical analysis and a multi classifier artificial neural network is trained with a separate data set to improve the rate of detection of the minority attacks. The accuracy of 99.34 % is recorded for minor attacks detection.